Add test to verify we can cancel the GetUserMedia request right after sending the request

chrome/test/functional/webrtc_call.py

Index: chrome/test/functional/webrtc_call.py
diff --git a/chrome/test/functional/webrtc_call.py b/chrome/test/functional/webrtc_call.py
index 68eec9ed2ceca4bdd09b2deaa2111123dce87dcf..bb431e2a8b81c7b33a779eaba42c9139e35ddaab 100755
--- a/chrome/test/functional/webrtc_call.py
+++ b/chrome/test/functional/webrtc_call.py
@@ -144,6 +144,18 @@ class WebRTCCallTest(pyauto.PyUITest):
     self.assertEquals('failed-with-error-1',
                       self._GetUserMedia(tab_index=0, action='dismiss'))
 
+  def testCancelGetUserMediaRequest(self):

[phoglund_chromium, 2012/07/04 14:20:58]

The first line in the docstring must not have any line breaks in it (it's sort
of the short description for the method). Maybe you can write something like
"""Check we don't crash on a quick reload after getUserMedia.

This test will quickly reload the page after running getUserMedia. This crashed
the browser before the fix for crbug.com/135043.
"""

[no longer working on chromium, 2012/07/05 14:05:19]

Done.

+    """Ensures the browser does not crash when cancelling a pending GetUserMedia
+    request. See issue 135043."""
+
+    url = self.GetFileURLForDataPath('webrtc', 'webrtc_jsep_test.html')
+    self.NavigateToURL(url)
+    self.AppendTab(pyauto.GURL(url))

[phoglund_chromium, 2012/07/04 14:20:58]

You did reproduce the crash with the test, right? Note that the javascript
calling here will mean that the reload will be noticeably delayed compared to
the bug's test page, since we need to go through the PyAuto test layers to call
your new reload function.

[no longer working on chromium, 2012/07/05 14:05:19]

Thanks for pointing this out. I just double checked, the crash I saw happened in
a later state, which hit a DCHECK in MediaStreamDeviceSettings::PostResponse(),
not exactly the same place as how it is described in issue 135043, but they
belong to the same issue.
And I also tried a bit on call getUserMedia and reload in sequence, but I could
not reproduce any crash, it looks like that if we call reload too soon, we are
able to catch up the request and avoid crashing.

+    self.assertEquals('ok-requested', self.ExecuteJavascript(
+        'getUserMedia(true, true)', tab_index=0))
+    self.assertEquals('ok-calceled-request', self.ExecuteJavascript(
+        'CancelGetUserMediaRequest()', tab_index=0))
+
   def _GetUserMedia(self, tab_index, action='allow'):
     """Acquires webcam or mic for one tab and returns the result."""
     self.assertEquals('ok-requested', self.ExecuteJavascript(